AI Providers

Eroteme AI uses five different AI providers across eight specialised agent roles. Each provider was chosen for specific strengths that match its role in the pipeline.

Provider Roles

Claude (Anthropic)

Roles: Base Rate Analyst, Bull Case, Meta-Judge

Claude handles three critical roles in the pipeline. As the Base Rate Analyst, it computes historical frequencies while deliberately ignoring current news to avoid recency bias. As the Bull Case analyst, it constructs the strongest argument for YES. As the Meta-Judge, it uses Extended Thinking to synthesise all 8 agent outputs into the final calibrated probability.

ChatGPT (OpenAI)

Role: Bear Case Analyst

ChatGPT is assigned the Bear Case — constructing the strongest argument for NO. Having a different provider argue against the bull case ensures genuine adversarial analysis rather than a single model debating itself.

Gemini (Google)

Roles: Data Analyst, Evidence Synthesiser

Gemini handles data-heavy roles. As the Data Analyst in the research layer, it pulls current statistics, historical precedents, and quantitative data. As the Evidence Synthesiser in the analysis layer, it provides balanced, objective evidence weighting across all research findings.

Grok (xAI)

Roles: Sentiment Analyst, Contrarian

Grok leverages its connection to the X/Twitter platform. As the Sentiment Analyst, it analyses real-time social sentiment around the event. As the Contrarian, it challenges consensus by identifying blind spots and scenarios where the majority view might be wrong.

Perplexity

Role: News Researcher

Perplexity's real-time web search capability makes it ideal for gathering the latest news and developments. It grounds the entire pipeline in current events, ensuring the analysis isn't limited to training data.

Provider Architecture

Each provider serves a distinct analytical function — no two agents share the same role. This ensures genuine diversity of perspective rather than the same model giving slightly different answers.

Research Layer          Analysis Layer          Meta-Judge
├─ Perplexity (News)    ├─ Claude (Bull)         └─ Claude (Synthesis)
├─ Grok (Sentiment)     ├─ ChatGPT (Bear)
├─ Gemini (Data)        ├─ Gemini (Evidence)
└─ Claude (Base Rate)   └─ Grok (Contrarian)

Standard Tier

For Standard tier predictions ($1 USDC), a single model runs the analysis. You can get a Standard analysis from Claude, ChatGPT, Gemini, or Grok. The model selection depends on the market category and availability.

Data Sources

In addition to the AI models, the pipeline draws on: